Page 1 sur 13

I would be grateful for a professional translation or correction of the following text. Please E -Mail to
DL2YEO@qrp4u.de

250 W S.M.P.S. with Power FETs

Safety Instructions

Caution mortal danger: The following circuit operates at a mains voltage of 230 Vac.
Because of rectification some of the components conduct dc voltage of more than 322 V.
Work has to be carried out only if the circuit is disconnected from the mains and de-
energized. Note that capacitors located to the primary side can be charged with high
voltage for several seconds even after switching of the mains voltage.

Experiments with the PC power supply encouraged me to an "advancement " design. The
new power supply is also a forward converter with a half-bridge topology. The difference
to the modified PC power supply consists in the following items: Two power FETs instead
of bipolar transistors used as power switches, a synchronous rectifier on the secondary
side instead of power diodes, the lack of a driver stage (current-proportional control) and a
more simple over-current and over-voltage monitor. With the magnetic components
(output transformer, driver transformer, chokes...) stripped from a PC power supply, the
new power supply delivers a max. power output of 250 W, whereby the efficiency
amounts up to 90 %. The power supply can handle 20 % over-load for a short duration.

Magnetic components from a PC power supply

The magnetic components of PC power supplies for AT boards do not indicate large
differences. They are usually designed for a switching frequency of 25... 40 kHz and a
power output of 200... 240 W. The transformers on the S.M.P.S circuit boards are to be
found often in a small or somewhat larger size. I am not able to say, whether the larger
size brings more power or if it is only an older design. For the new power supply I
preferred the larger transformers, because of more space for additional turns available at
all three transformers. The smaller transformers are completely filled with copper and
isolation material and therefore only conditionally suitable for a modification.

Fig. 1: Transformers from the PC power supply

Mains rectifier and filter

This circuit section is uncomplicated. The current compensated choke Dr1 (mains filter) is
followed by a NTC resistor for limiting the inrush current. Its cold resistance amounts to 5
ohms and after few minutes the warm resistance is less than one ohm. The 230 Vac

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 2 sur 13

rectifier is generously dimensioned with 4 A and hence no cooling is necessary. A

criterion for the selection of the capacity of C3 and C4 is the height of the admissible
ripple voltage Ubr and the number of mains voltage half waves to be bridged. For Ubr =
25 V and zero half waves two 470 uF capacitors in series are sufficient. This specification
applies to maximum load during low mains voltage Umin = 230 Vac - 15%

Fig. 2: Filter, rectifier and power switches

Power switches

As power switches FETs were used due to their short rise and fall times and the easy,
component-saving driver circuit. If one is content with switching times of 100 ns, a small
driver transformer and two gate resistances are sufficient for triggering the FETs. Re-
dimensioning of the secondary winding for the supply of the gate voltage is not to be gone
around unfortunately. The single turn and 2 x 8 turns must be removed from transformer
Tr4. Instead of this 2 x 16 turns (bifilar) have to be applied. With a winding ratio of 16 : 26
and a 20 Vs control signal from IC1 the FETs gate voltage is up to 10 V high. With this
voltage one achieves the indicated FETs Ron of 0,75 ohms and thus very small
conduction losses. Dynamic (switching) losses at 50 kHz and the before mentioned
switching times assume negligible. The PWM IC drives enough current for fast on /off
switching the FETs. Increasing the switching frequency from before 33 kHz (PC power
supply) to 50 kHz (new power supply) allows to transfer more energy via the transformer.
You can not increase the latter with a given transformer whatever you like, since the
transformer is usable for a certain frequency range only. Attempts showed that the
transformer can cope with the factor 1.5 without problems (overheating).

Control circuit

After switching on the 230 Vac mains voltage an auxiliary voltage is available from the
small 50 Hz transformer Tr1 for starting up the PWM controller SG3525. The P acting
operation amplifiers in the SG3525 compares a portion of the 13,8 output voltage (actual
value) with the internal +5,1 V reference voltage (set value) and forms from it the
correction variable for the pulse width modulator. The modulator sends alternate control
pulses via its two outputs to the transformer Tr4. The pulse duration is reciprocal to the

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 3 sur 13

correcting variable. Increased loading to the +13.8 V output makes for wider pulses,
lighter loading causes narrower pulses. The switching frequency of the power switches is
50 kHz. For higher frequencies the FETs are usable but not the magnetic components
that were taken over by the PC power supply. The oscillator frequency is determined by
the components attached to pin 5 and 6. R14 determines the dead time, which is
absolutely necessary to avoid two switching transistors conduct at the same time. Due to
the not present storage time for FETs a very small value could be set. With 1 us deadtime
and 20 us period duration the FETs can theoretically lead current for 95 % of the time and
thus deliver energy to the output. Charging C13 after switching on causes a soft start with
narrow pulse first and wider control pulses afterwards. Terminal (a) of the driver
transformer Tr4 remains free. Only one half (26 t) of the primary turns (b - c) and the 16
turns of the secondary winding are sufficient to form the necessary ratio of 0,6.

Fig. 3: PWM control and monitoring

Monitoring functions

Two protection circuits are included in the new power supply. The transformer Tr2 is used
as current detector and produces at R16 a voltage that is proportional to the current flow
through the power switches. If the voltage at the shutdown pin 10 exceeds the limit value
adjustable with P1, the control IC switches off immediately and restarts after a short
duration. The reason for this is usually a too high current at the secondary side of the
power transformer, caused by a short-circuit or an overload to the output. The load and
the circuit itself are likewise protected from overvoltage at the Vo output. The SG3525
switches off at Vo > 15 V. Note: Both protection circuits are ineffective if the slider of P1 is
adjusted to Gnd potential.

Synchronous rectifier

One must deal with a power dissipation up to 17 W at 18 A output current when using a
rectifier with fast recovery diodes. With a 30 A / 45 V Schottky diode the looses are 12 W
nevertheless. This rate forms the highest proportion referring of the total losses. The
losses at the mains rectifier, power switches, transformers and the output choke are
together below this value.

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 4 sur 13

An improvement brings here a semi synchronous rectifier with two low impedance power
FETs. FETs with a low Ron of e.g. 15 mOhm have only a voltage drop of 0,3 V at 18 A
during the conduction phase. Good Schottky diodes are rated at 0,6 V. In the literature
however with push-pull topology such circuits are advised against. As reason it is
indicated that the choke current flows in opposite direction (revers mode) through the
parasitical body diodes of the FETs as long as they are in the off state. High switching
losses occur due to the storage charge of the body diodes which has to be removed first
during the transition to the normal operation. This looses destroy the benefits achieved
during the conduction phase.

The following circuit avoids this disadvantage, since the body diodes do not operate in the
reverse mode. The free wheeling diode D3 takes over choke current since it has a
substantially smaller forward on voltage Uf compared to the FETs body diode. D3 is from
Schottky type and is arranged before the choke Dr2. The diode does not have a storage
charge and thus switching on/off is extremely fast combined with low losses. It was
removed once for by way of trial. The FETs heat sink warmed up thereupon by around
+10 C, although the body diode of the used IRFZ44 has already a very good trr (reverse
recovery time) of 47 ns.

With a typical duty cycle of 57 % the losses in the two FETs make together 3.6 W. The
free wheeling diode D3 is conducting for the remaining time and produces 4.6 W. Less
than 8.2 W can only be attained, if one replaces the free wheeling diode also by a FET.
Since triggering this FET is more complex than from VT3 and VT4, I did without this
measure. A second reason is that with mains undervoltage or high output current the
switch-on time of VT3/4 and not the switch-on time of D3 rise.

Fig. 4: Synchronous rectifier

Construction and alignment

For the power supply assembly a glass epoxy circuit board with the dimensions 82 x 122
mm is needed. One should not use another material. It applies to fasten relatively heavy
components and realize copper tracks that are able to carry high current. The
components for regulation and monitoring are mounted on a small strip board. Sorry, but I
was too lazy to design a PCB layout for this circuit part.

Fig.5: PCB layout (1:1) and assembly

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 5 sur 13

For the interconnection of the components for regulation and monitoring a 40 x 45 mm

small strip-board is sufficient. The copper tracks (pink) are to be removed in the indicated
places. A wood or a metal drill with a diameter from 3 to 4 mm is suited best tool for this
work. Cable links are drawn in as broken lines. They are forgotten fast with the assembly.
The same applies to the horizontal ground potential bar within the upper area of the strip -
board that distributes Gnd potential onto the vertical copper tracks.

Fig.6: View to the strip-board soldering side (2:1)

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 6 sur 13

Fig. 7: View to the strip -board component side (2:1)

Transformers

The following drawing shows the transformers stripped from a PC switching power supply.

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 7 sur 13

The data were determined so far as possible by measurements, counting turns and
calculations.

Before using the transformers it has to be checked exactly whether the size, number of
layers, wire size, number of turns and phasing correspond to the specification in the
drawing and the photos. If doubts exist in the matching, the transformers should better not
be used.

Fig. 8: PC transformers and modifications

Heat sinks

At the heat sinks is nothing special. They are manufactured from an approx. 1 mm thick
aluminum plate. VT1 and VT2 are to be fastened isolated on the first heat sink. The FETs
may not have a electrical connection against each other and against the heat sink. With
professional assembling of the transistors touching the heat sink is without danger. On the
secondary side it is somewhat simpler. VT3, VT4 and D3 carry no dangerous voltage and

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 8 sur 13

need thus no isolation for this reason. Because of the FETs case and the Schottky diodes
case have the same potential, there is no objection against mounting all three
components directly onto the second heat sink. It is to be made certain however that there
is not electrical connection between the heat sink itself and the power supply housing or
electrical components.

Fig.9: Heat sink drawings

Parts list (1)

Resistors, capacitors and semi conductors

Parts No. Value

R1, 2 120 kOhm, 0,5 W
R3 100 Ohm , 2 W
R4, 5, 9 1 kOhm
R6 10 Ohm, 2 W
R7, 10 10 kOhm
R8 1,5 KOhm + 150 Ohm
R11 5,6 kOhm
R12, 13, 14 47 Ohm
R15, 16 150 Ohm
P1 10 kOhm trimming pot, 10 turns
NTC Heissleiter, 5 Ohm at 25 C
C1, 2 0,1 uF 250 Vac
C3, 4 470 uF 200 V, 22 x 36 mm (diam. , H)
C5, 15 2,2 nF
C6 1 uF, 250 Vac
C9, 10 2200 uF, 35 V low ESR, 16 x 34 mm (Diam., H)

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 9 sur 13

C7 100 F, 35 V
C8, C20 10 nF
C11,12 0,22 F
C13 10 uF, 25 V
C14 2,2 nF Styroflex
C16 2,2 uF
C17, 18, 19 0,047 uF
D1, 2 PXPR1507 etc. fast 200 V / 1A diode
D3 MBR3045, 30 A / 45 V Schottky diode
D4, 5, 6 BAT 46
D7 Zener diode, 13 V / 0,5 W
D8 1N4148
VT1, 2 IRF730
VT3, 4 IRFZ44N
IC1 SG3525A
Gl1 Rectifier bridge, dual in -line B40C800 DIP
Gl2 Rectifier bridge 400 V / 4 A

Parts list (2)

Transformers, chokes and miscellaneous

Parts No. Value

0,5 W print transformer EE20/10, 15 Vac at 34 mA,
Tr1
24 x 32 mm (Reichelt/Conrad)
16 x 15 x 5 mm (W,H,D)
Tr2 1 Wdg. prim.
2x 100 turns sec.
40 x 35 x 12 mm (W,H,D) e.g. Tokin 25812 or.
25801
2x 20 turns prim. (L = 7 mH between a <=> c)
Tr3
2x (3 + 4) turns sec. (L = 200 uH between d <=> f or
d* <=> f*)
2x 4 turns sec. auxiliary winding for driving VT3/4
22 x 19 x 6 mm (W,H,D)
Tr4 2x 26 turns. prim.
2x 16 turns. sec.
Dr1 current compensated 2A mains voltage choke
20 uH, T26-106 (yel. / white), 16 turns. 2x 1 mm Cu
wires in parallel
Dr2
better Magnetics Kool 259-77934-A7, 20 turns. 2x1
mm Cu wires in parallel
Additional mains general purpose 230 V / 2 A
filter

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 10 sur 13

Si 3,15 AT fuse, slow blow

PS Two pole mains switch
PCB, heat sinks, isolation material, heat sink
Miscellaneous
compounder etc.

The grey marked cells indicate the components, which can be taken over by a PC power
supply. The electrical data must be compared before using them and the indicated
modifications have to be executed.

Testing the power supply

I urgently advise against immediate connection to 230 Vac. Testing of the new power
supply should take place in several test phases for safety reasons and for the avoidance
of component destruction. The high voltage causes an immediate destruction of the
components in the case of an error in the circuit.

Warning: Check temperature of components only if the mains voltage is switched off.

Phase 1:The first test applies to the PWM-IC and the power switch control. For running
the PWM-IC a lab power supply adjusted to 24 Vdc must be attached to Gnd and the
positive plate of C7 (Vx). After switching on the the IC generates sharply rising and falling
control pulses with maximum pulse duration at the output pins 11 and 14. With an
oscilloscope measured signals at the gate of VT2 (VT1) must look like the one shown in
figure 9. It is very important that the signals have the indicated shape, voltage and
frequency. The signals at the gates must be further in opposite phase against each other.
Otherwise, both FETs would conduct at the same time and produce a short-circuit when
later applying supply voltage .

Fig. 9: VT2 (VT1) gate-source voltage

Phase 2: Now, connect three car light bulbs (12 V / 21 W) to the 13,8 V output terminals.
A 48 V / 1 A mains transformer feeds the S.M.P.S. via the L1 and N terminal with a
galvanical isolated Ac voltage. The + 24 Vdc lab supply is still connected during this test.
60 Vdc at C3 / C4 is in Europe defined as a non-dangerous voltage rate. At this voltage

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 11 sur 13

rate the switching transistors can start operating and one can perform test without danger.
For measurements with a dual-channel oscilloscope Gnd from the secondary section has
to be connected temporary to the (Y) test point of the primary section with a wire link. The
bulbs glow at Vout = 4,3 Vdc if everything is right. Rectification is executed by the FETs
body diodes only, because the VT3 and VT4 gate -source voltage is not high enough to
switch on the FET. The PWM controller tries to offer 13,8 V at the output at maximum
pulse duration. The later cannot be successful due to the low 60 Vdc input voltage and the
present transformer ratio.

Fig. 10: Voltage at test point (X) against (Y) und cathode D3 against Gnd

Phase 3: If everything is all right up to now, one can proceed with the exciting test at 230
Vac. The laboratory power supply, the 48 V transformer, the measuring instruments and
all provisional cable links attached for the test etc. must obviously be removed. The three
car bulbs are further needed as a load and for the functional checks. If after applying of
the 230 Vac mains voltage the lamps light up brightly, the output voltage amounts to 13.8
V and no undefined noises or smells are noticeable one has won the first round. If a non
recognizable error has passed the pre-testing the two switching transistors and copper
tracks say good-bye with a more or less loud bang. With 5,7 A the duty cycle D = tp /
T = 5 us / 10 us is approximately 50 %.

Phase 4: For the following load test a dummy load is needed that can handle up to 300 W.
Because such high power resistors are not laying in the junk box and purchasing them is
very expensive, I took instead a 50 m a ring installation cable (3 x 1.5 mm2) . An
individual wire has a resistance 0.6 ohms and can dissipate the mentioned watts without
problems. Depending upon interconnection of the three wires, load resistances of 0,6 / 1,2
and 1.8 ohms are realizable. By the ampmeter impedance, including the appropriate
measuring wires positioned in series, the resistance value increases by approx. 0.1 ohms.
At Vo = 13,8 V the following table indicates the power output Po and the current Io as a
function of the loading.

Rl [Ohm] car bulbs Io [A] Po [W]

1x 12 V / 21
-/ - 1,9 26
W
2x 12 V / 21

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 12 sur 13

-/- W 3,8 52
3x 12 V / 21
-/- 5,7 78
W
1,8 + 0,1 -/- 7,26 100
1,2 + 0,1 -/- 10,6 146
2x 12V / 21 10,6 +
1,2 + 0,1 198
W 3,8
3x 12V / 21 10,6 +
1,2 + 0,1 224
W 5,7
0,6 + 0,1 -/- 19,7 270

Additional measures for RFI noise reduction

Experience during the PC power supply modification have shown that the filtering on the
PCB is not sufficient for amateur radio application. A pre-arranged general purpose mains
filter and a home made Pi filter direct to the 13,8 V output are used for improved RF noise
reduction. To favor of the control loop stability the PI filter voltage drop is not eliminated.
Output voltage changes of several ten millivolts at load changes have no importance for
running a 100 W transceiver. The additional filter have to be mounted inside the S.M.P.S.
case very close to the cable inlet and outlet.

Fig. 11: External components for RF noise reduction

Operation experiences

Up to 10 A continuous output current or operation with 50 % ESD and 18 A peek current

one can do without a fan if sufficient natural air flow is present and the ambient
temperature does not exceed 30 C. A small CPU fan (40 x 40 mm) should be used for
more than 10 A continuous current. The heatsink surface is not large enough to keep the
FETs junction temperature below the limit value ( Tj < 100 C). With a CPU fan the heat
sink temperature remains below 28 C (Tu = 20 C). The following table shows the
measured and calculated power dissipation Pv of the basic components at 250 W output
power.

Abbr. Parts Pv [W]

Gl2, Dr1 Mains rectifier and filter 2,5
VT1 4,0
Switching transistors
VT2 4,0

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003
 Page 13 sur 13

R3 / C5 Snubbers 1,5
Tr3 Output transformer 2,0
VT3 1,8
Synchronous rectifier
VT4 1,8
D3 Free wheeling diode 4,6
Dr2 Choke 2,0
IC1 PWM control circuit 1,0

Sum 25,2

Back to the menu

file://C:\DOCUME~1\bada\LOCALS~1\Temp\GLO47MJA.htm 31/08/2003